FormulaBOINC

[Ken g6]

For now we're #1 in the sprint, and #2 overall.

 

[StefanR5R]

 

[Orange Kid]

A bit early, but here is day two.
We have captured first in the sprint and are stretching our lead in the marathon.
The big race seems to be for first, will we hang on? Tune in tomorrow to find out.

Yafu..........................Sprint.............Day Two
1 TeAm AnandTech .....................25 ...........2,090,393
2 OcUK - Overclockers UK ..........18 ...........1,855,630
3 Crunching@EVGA ....................15 ..............893,201
4 Overclock.net .............................12 .............474,302
5 meisterkuehler.de Team ............10 .............240,931
6 Portugal@Home ..........................8 ................9,128
7 AMD Users ...................................6 ...................78
8 Sicituradastra. ..............................4 ...................37


Yafu..........................Sprint.............Day One
1 OcUK - Overclockers UK ..........25 .............684,215
2 TeAm AnandTech ...................18 .............655,914
3 Crunching@EVGA ....................15 ............382,135
4 Overclock.net ............................12 ............151,280
5 meisterkuehler.de Team ...........10 .............110,658
6 Portugal@Home ........................8 .................6,403
7 AMD Users ................................6 .....................78


Yafu..........................Marathon
1 TeAm AnandTech .....................25 ............6,834,633
2 Crunching@EVGA ....................18 ............3,492,478
3 OcUK - Overclockers UK ...........15 ............2,900,065
4 meisterkuehler.de Team ............12 ............2,376,687
5 Overclock.net .............................10 ............1,662,215
6 Ars Technica ...............................8 .............1,007,561
7 The Scottish Boinc Team ............6 ................602,777
8 AMD Users ..................................4 ................347,769
9 Portugal@Home .........................2 .................254,703
10 SwissTeam.NET .......................1 ...................78,623

 

[yodap]

Nice job over here! Where the heck are pulling those Yafu points out of?

 

[StefanR5R]

Speaking for myself only of course, but... I assure you I rinse these points carefully before delivery, and wash hands too.

 

[crashtech]

I have concluded that YAFU is a dumb project for a Sprint, especially for the wide range of machines I have. Presumably I could set different location preferences so my 4T CPUs could get 4T WUs, my 8T machines could get 8T WUs etc., but it shouldn't be that hard; some awareness by the app about how many cores are available on a machine while doling out work would be much more efficient and ensure that the project would get more cycles.

Edit: It's been so hard for me to keep 100% utilization on YAFU that I'm allowing other projects to fill in the dead spots. Maybe configuring some locations is mandatory to do well.

 

[GLeeM]

Some of my WUs run for about 24 hours and then pause to let the next WU run.

How do I get the manager to finish long WUs and not pause/switch?

If I leave blank "Switch between tasks every ______ minutes." will it never switch? I have it set to 660 minutes now.

 

[StefanR5R]

@crashtech, the "YAFU" and "YAFU for small composites" applications are variably-threaded. At the day before the race there weren't enough tasks for them though. But since the race began, I have them running on one 6C/12T node. I haven't monitored this node closely whether it ever ran out of tasks; I believe it didn't.

 

[crashtech]

@crashtech, the "YAFU" and "YAFU for small composites" applications are variably-threaded. At the day before the race there weren't enough tasks for them though. But since the race began, I have them running on one 6C/12T node. I haven't monitored this node closely whether it ever ran out of tasks; I believe it didn't.

I didn't know that, only that the "16T" ones seem variable. Maybe switching to just the three will help, leaving the 4 and 8T ones to those who want them. The 4T WUs in particular seem problematic for me, I can't get two of them to run on an 8T machine.

 

[StefanR5R]

Tomorrow is going to be a thrilling day. Which of all those tasks that are already running for 10, 20, 30, 40 hours (and had their progress bar at 100 % for ages already of course) will finish before the sprint ends?

And of the longer-running tasks that finish in time, which ones will get fair credit, and which ones will end up with poor credit?

I think I better keep myself far away from my hosts tomorrow; this may all be too much excitement.

 

[ao_ika_red]

Running YAFU on old platform really show the significance of architectural improvement in recent decade. I've got 2T WU that needed 54 hrs to be completed but only got 2k credits from YAFU.
https://yafu.myfirewall.org/yafu/result.php?resultid=1180168

 

[StefanR5R]

Here are two random results from hosts with same processor, same OS, same application (yafu-8t).

Sent                        Time reported               Run time (s)   CPU time (s)    Credit   
16 Sep 2017, 15:51:42 UTC   16 Sep 2017, 16:41:17 UTC     2,975.00      15,970.08      935.31
16 Sep 2017, 15:48:33 UTC   16 Sep 2017, 16:30:26 UTC     2,513.00       7,035.51    2,508.59

YAFU rhymes with SNAFU.

 

[TennesseeTony]

Wow, points are totally based on a random number generator.

 

[TennesseeTony]

For the year, [H] in league 3 is now ahead of Gridcoin in League 1 in total points!

 

[ao_ika_red]

Here are two random results from hosts with same processor, same OS, same application (yafu-8t).

Sent                        Time reported               Run time (s)   CPU time (s)    Credit   
16 Sep 2017, 15:51:42 UTC   16 Sep 2017, 16:41:17 UTC     2,975.00      15,970.08      935.31
16 Sep 2017, 15:48:33 UTC   16 Sep 2017, 16:30:26 UTC     2,513.00       7,035.51    2,508.59

YAFU rhymes with SNAFU.

After reading this, I smell vaporised tears from my laptop. I think I will revert it back to seti machine (seti's stock apps fare better on the C2D than Kaveri's Athlon) after the sprint over.

 

[StefanR5R]

@ao_ika_red, the difference of credits/h from my good and my bad sample to your sample is quite shocking indeed, even considering thread count (8 vs. 2). I wonder what's going on there, can't quite believe that this is entirely due to processor architecture.

My two samples are from 64 bit Linux, but from a quick glance, my 64 bit Windows results are in a similar range.

BTW, according to the turbo clocks I am seeing, YAFU does not use AVX.

 

[ao_ika_red]

@StefanR5R to be honest, I don't know. I just found my laptop stashed in a bookshelf early this week. I just cleaned and changed the TIM. For a few days it run seti with stable clock @ 2.1 GHz / 65 C. I only run YAFU when it announced on FB sprint and the only validated major task so far is just that one.

Is there any possibility of bad drive may ruin the process? Does yafu need constant internet connection all the time?
Because it has pretty old drive and its wifi adapter is quite dodgy if I left it turned on for a while.

Based on yafu task report, my laptop consumed a lot of time in completing linear algebra subtasks. Does cpu cache size contribute to the slower process? Or maybe its integer performance is just not up there.

 

[StefanR5R]

Some of my WUs run for about 24 hours and then pause to let the next WU run.

How do I get the manager to finish long WUs and not pause/switch?

If I leave blank "Switch between tasks every ______ minutes." will it never switch? I have it set to 660 minutes now.

I haven't found upstream documentation of the default value yet. BOINC FAQ wiki mentions a default of 60 minutes at a few pages.

I have it usually set to something like 6000 minutes.

 

[Kiska]

I haven't found upstream documentation of the default value yet.

Yes the default value is 60 minutes between switching tasks. However that only switches tasks IF, you have tasks from more than 1 project in your queue. If you only have 1 project of tasks, then task switching does not occur.

You can find more about it here: http://boinc.berkeley.edu/wiki/PreferencesXml

<cpu_scheduling_period_minutes>60.000000</cpu_scheduling_period_minutes>

And a description of what it does here: http://boinc.berkeley.edu/wiki/Preferences

 

[Orange Kid]

A first place !!!!
Congrats to all on a fine race. Was a tuff one for a sprint.
A week and a half till the next.
We are in second overall with this and a couple of fresh point from somewhere.

Yafu..........................Sprint.............Final
1 TeAm AnandTech ...................25 ..........3,889,617
2 OcUK - Overclockers UK .........18 ...........3,097,330
3 Crunching@EVGA ...................15 ...........1,275,313
4 Overclock.net ...........................12 ..............754,566
5 meisterkuehler.de Team ..........10 ..............327,528
6 The Scottish Boinc Team ..........8 ...............270,695
7 Portugal@Home .......................6 .................13,559
8 AMD Users ...............................4 ...................4,513
9 Sicituradastra. ...........................2 .......................37

Yafu..........................Sprint.............Day Two
1 TeAm AnandTech .....................25 ...........2,090,393
2 OcUK - Overclockers UK ..........18 ...........1,855,630
3 Crunching@EVGA ....................15 ..............893,201
4 Overclock.net .............................12 .............474,302
5 meisterkuehler.de Team ............10 .............240,931
6 Portugal@Home ..........................8 ................9,128
7 AMD Users ...................................6 ...................78
8 Sicituradastra. ..............................4 ...................37

Yafu..........................Sprint.............Day One
1 OcUK - Overclockers UK ..........25 .............684,215
2 TeAm AnandTech ...................18 .............655,914
3 Crunching@EVGA ....................15 ............382,135
4 Overclock.net ............................12 ............151,280
5 meisterkuehler.de Team ...........10 .............110,658
6 Portugal@Home ........................8 .................6,403
7 AMD Users ................................6 .....................78

Yafu..........................Marathon
1 TeAm AnandTech ....................25 ...........8,688,751
2 OcUK - Overclockers UK .........18 ...........4,157,735
3 Crunching@EVGA ...................15 ...........3,875,605
4 meisterkuehler.de Team ..........12 ...........2,462,987
5 Overclock.net ...........................10 ...........1,959,805
6 Ars Technica .............................8 ............1,007,561
7 The Scottish Boinc Team ..........6 ...............854,417
8 AMD Users ................................4 ...............352,204
9 Portugal@Home ........................2 ...............259,134
10 SwissTeam.NET ......................1 ..................78,623

 

[TennesseeTony]

Well done TeAm!

 

[StefanR5R]

Congrats to the TeAm for doing this well in the Sprint + Marathon + overall, as well as for our 5%-RAC badge. And to @Orange Kid, @GLeeM, @crashtech and xii5ku for 25/5/1%-RAC badges. And to @ao_ika_red for coming in as 6th per RAC in the TeAm — see what a determined Core2Duo can still achieve even in a 3 days sprint with tasks that can take days themselves and against a screwy credit system.

 

[Orange Kid]

My E6850 thanks you for making it feel special

 

[Jondi]

Well done, we hoped to make more of a fight of it on the final day but you guys had too much for us and/or managed to get the best out of the credit server.

Couldn't be closer now in the overall standings, see you at the next one.

 

[StefanR5R]

Thanks @Jondi. Your group has been a very tough opponent again.

That capricious deity that hands out yafu credits looked mildly on the results of some (but not all) of our team mates.